Abstract: A new blind intelligent image watermarking technique proposed in this paper adopts intelligent information disposal-independent component analysis. The characteristics of the human visual system are incorporated into the watermark embedding and a closed-loop embedding in discrete wavelet transform domain is proposed. The watermark is rearranged by chaotic before embedded, and it can be detected and extracted correctly not merely be detected without any information about the original image, watermark and attack. It can also extract the multiple embedded watermarks. The accuracy of watermark extraction depends on the key and the statistical independence between the original images and watermark. Experimental results demonstrate that the proposed intelligent watermarking technique is robust with respect to traditional image processing including cropping, filtering, image compression and geometric distortions such as rotation, scaling. It even has a good robustness against combination of several attacks. All attacks are produced by popular watermark test software-Stirmark.